What a voice clone needs
Percify’s guidance for cloning yourself is clear audio without noise, a natural pace, and 5 to 60 seconds; for a saved avatar voice it asks for 10 to 30 seconds of clean speech. The checker measures each of those:
| Check | Good | Fix if not |
|---|---|---|
| Length | 10 to 30 seconds of speech | Under 6 seconds is too short; over a minute, trim to the cleanest part |
| No distortion | Never hits full scale | Move back from the mic or turn the input down |
| Loud enough | Speech above about minus 28 dBFS | Move closer or raise the input level |
| Background noise | Voice 30 dB or more above the room | Record in a quiet room away from fans, traffic and echo |
| Mostly speech | Talking through most of it, no long silences | Talk continuously and cut pauses over 3 seconds |
- These are our guidance, not hard limits: a sample marked could be better still clones, just less closely.
Recording tips
- Use the quietest room you have. Soft furnishings cut echo.
- Keep a hand’s width between your mouth and the mic, a little to the side.
- Speak the way you want the clone to sound: the same energy and pace.
- Record your own voice, or a voice you have permission to clone.
Private by design
The recording is decoded and measured with the Web Audio API in your browser. It is not uploaded, and the recorder turns off echo cancellation and noise suppression so the check hears what a clone would hear.
Checked on 17 September 2026 against Percify’s voice cloning guidance and tests on sample recordings.